CoChallenge | Devpost

Introduction

We were inspired by the gap between learning and proving skill in the tech industry. Many learners strive to demonstrate real-world ability, while recruiters seek resumes that reflect project-based skills. We wanted to create an educational platform where learning connects to opportunity—and where talent can be discovered through meaningful projects.

What it does

CoChallenge is a platform that connects learners and recruiters through project-based learning and open-ended challenges. Learners can generate personalized project ideas, and agentic AI transforms those ideas into structured, step-by-step learning paths. Once users build confidence, they can tackle real challenges posted by companies. Submissions are company-evaluated, and users earn an Elo-based ranking that reflects their demonstrated ability, helping recruiters identify top talent. Users can export their projects to GitHub for further building.

How we built it

We built CoChallenge as a full-stack web application. On the frontend, we focused on a clean, intuitive interface built on React and Tailwind for generating ideas, following guided lessons, and submitting challenge work. On the backend, we integrated AI agents controlled by Google's ADK orchestrated on a Flask server to dynamically create structured project plans and learning scaffolds. We also implemented a ranking system inspired by Elo ratings to track user performance and progression based on company evaluations. MongoDB and Supabase were used to store data and authenticate users.

Challenges we ran into

One of the biggest challenges was designing AI-generated lesson plans that are both structured and adaptable to different skill levels. We also had to think carefully about how to fairly evaluate open-ended submissions and translate that into a meaningful ranking system. Ensuring a smooth user experience across both learners and recruiters—two very different user groups—was another key challenge. Safely handling the output from agents was another challenge, and required careful parsing and correction tools to save to our database.

Accomplishments that we're proud of

We’re proud of creating a system that doesn’t just teach—but verifies skill through real work. The integration of agentic AI to generate actionable learning paths, combined with a recruiter-facing challenge system, makes CoChallenge feel like a complete ecosystem rather than just a learning tool. The Elo-based ranking adds a competitive and merit-based layer that encourages continuous improvement.

What we learned

We learned how powerful it is to combine AI with structured learning design. Building CoChallenge also taught us a lot about balancing automation with user control—especially when generating educational content. Additionally, we gained insight into designing systems that serve multiple stakeholders, each with different goals.

What's next for CoChallenge

Next, we want to improve the evaluation system by incorporating more nuanced feedback and potentially hybrid AI + human grading. We also plan to expand the range of challenges, onboard real companies, and refine the ranking system to better reflect skill across different domains. Ultimately, we envision CoChallenge becoming a go-to platform where learning, building, and hiring all happen in one place.